ChatGPT can pass the bar exam. Does it really count?

Back when I was studying journalism, we had an assignment called News Day, designed to replicate a day in the life of a journalist. You arrived at school in the morning and you were assigned a story to categorize by the end of the day. I forgot what my specific story assignment was – it was 12 years ago – but it had something to do with climate change. What I remember, with painful lucidity, is an interview with an academic who had agreed to help me.

After about 10 minutes he correctly guessed from my questions that I didn’t understand the problem, whatever it was. He told me to call him back after doing more research. It’s been 12 years and I still remember this incident every time I conduct an interview. Something like this hasn’t happened yet, probably because fear of repetition compelled me to do more than the minimum amount of research possible.

Journalism units made up a third of my communications degree. These units have been designed to be as practical as possible, to mimic real journalism. In addition to being fully owned by this environmental science professor, I remember being given practical advice on who does and does not make a good source, how to structure long feature articles and how to present articles to a publisher.

On the other hand, I remember almost nothing from my communication classes, which consisted of impractically dense texts and essays on topics such as “language and speech.” It’s a blur.

ChatGPT would have easily passed these communication courses. On Tuesday, OpenAI announced that, thanks to a new update, its The ChatGPT chatbot is now quite smart to not only pass the bar exam, but also rank in the top 10%.

It is easy to react to this with fear or dread. I certainly had a moment of undirected terror. But ChatGPT being technically qualified to practice law is not necessarily a sign of the AI ​​apocalypse. It may even be a good thing.

It helps to remember what the premise of a large-language AI model like ChatGPT is. It is an artificial intelligence trained on huge troves of data, from web pages to scholarly texts, which then uses a complex predictive mechanism to generate human-sounding text. It’s amazing, and ChatGPT is awesome to be able to take on such a high level challenge as the bar exam. But that’s not the paradigm shift it may sound like.

Exams are performances. They don’t test how knowledgeable students are, they test how much knowledge students can cram into their brains and regurgitate for a few hours. Much of this information is then forgotten, often happily. Exams are closer to tests of dedication than knowledge. Of course, humans can’t compete with software when it comes to aggregating information. But that is not the purpose of a review.

it’s so funny to me that the AI ​​people think it’s awesome when their programs pass a test after being trained on all the answers

—Paris Marx (@parismarx) March 15, 2023

It’s a new world. Goodbye homework!

— Elon Musk (@elonmusk) January 5, 2023

Exams as a format are still relatively safe, especially if schools return to the good old pen and page. Homework and college essays are on more fragile ground.

ChatGPT has scared many educators. Students in New York and Los Angeles are barred from using the app, as are children in many Australian high schools. Prestigious universities, including Oxford and Cambridge, have also banned your friendly neighborhood chatbot.

Educators who worry about new technologies are an old story. In the 1970s, some calculators feared they would ruin math, but actually made it easier to teach more complicated equations. An AI like ChatGPT can do a lot of homework, but so can Google. AI is flashier and more sophisticated, but it’s not a new threat. Teachers now tell kids not to cheat using ChatGPT, just like teachers told me not to use Wikipedia.

I did not listen. I still used Wikipedia. I still use Wikipedia. Every day.

Asking me not to was unrealistic, just as it is unrealistic to expect students not to take advantage of ChatGPT and the flood of services that are about to be launched by Google, Microsoft and Meta. Even if such a rule could be applied, it would be counterproductive. If AI is to be a part of life, it’s best for students to find the best way to work with it rather than against it. We all have a little Luddite spirit within us, but the real struggle of the Luddites has been futile.

After years of AI hype, ChatGPT has reinforced the idea that AI is likely to disrupt many industries. But disruption can mean change rather than destruction. In many cases, this change will be for the better. Education is a prime candidate for improvement. The industry is notoriously slow and can be pushed by AI without the risk of mass layoffs, as teacher jobs are generally more secure than many others.

The question is whether high schools, colleges and universities will adapt to the challenge. Elon Musk tweeted that AI can usher in the end of homework. Maybe. What if education adapted? Essays can be replaced with presentations or practical tasks relevant to the area of ​​study. In the remaining essay assignments, students may be graded more on persuasiveness rather than a basic ability to convey course concepts.

If university courses, for example, cannot be made more practical, it may be a sign that they never taught anything practical in the first place.

If I could time travel and use ChatGPT in my college course, it would only help me cheat on assignments that were the least educational in the first place. It wouldn’t have made a difference to the journalism assignments that were actually useful – the ones that taught me to prepare for interviews, always use the rule of three, and end commentary articles with a wink. look at the opening paragraph.

Editors’ note: CNET uses an AI engine to create personal finance explanations that are edited and verified by our editors. To learn more, see this article.